|Publication number||US6950977 B2|
|Application number||US 10/097,609|
|Publication date||Sep 27, 2005|
|Filing date||Mar 15, 2002|
|Priority date||Mar 15, 2001|
|Also published as||US20020166095|
|Publication number||097609, 10097609, US 6950977 B2, US 6950977B2, US-B2-6950977, US6950977 B2, US6950977B2|
|Inventors||Yoav Lavi, Alon Boner|
|Original Assignee||3G.Com, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (7), Non-Patent Citations (2), Referenced by (20), Classifications (8), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is an application filed under 35 U.S.C. § 111(a) claiming benefit pursuant to 35 U.S.C. § 119(e)(1) of the filing date of the Provisional Application 60/275,674 filed on Mar. 15, 2001 pursuant to 35 U.S.C. § 111(b).
The present invention relates generally to the field of error detection and correction for digital communications and more particularly to the area of turbo-coding and cyclic redundancy checking (CRC).
A known algorithm used for the correction of errors when data is transmitted in a noisy environment is known as turbo-decoding. In turbo-decoding, data bits to be transmitted are divided into block, and for each block, special bits are added and transmitted along with the data bits. The special bits that are added to each block enable the recovery of the original data, which is most beneficial in situations where the transmission is prone to noise that can corrupt the data when transmitted. The basic concepts involved in turbo-decoding are described, for example in Berrou, C., Glavieux, A. and Thitimajshima, P., Near Shannon Limit Error-Correcting Coding and Decoding: Turbo Codes, ICC, pp 1064-1070, 1993; and Berrou, C et. al., Turbo Codes: General Principles and Applications, Proceedings of the 6th Tierrenia International Workshop of Digital Communications, Tierrenia, Italy, September 1993, as well as in U.S. Pat. No. 5,446,747, all of which are incorporated herein by reference for all purposes.
The basic operation of a conventional turbo-decoder is discussed at length in the references mentioned above. Each of the conventional methods related to turbo-decoding discussed in the references mentioned above is based on several iterations performed on each block of data. As each iteration is conducted, gradually more and more errors are corrected until no further errors remain. In some real-time systems, the time dedicated to perform the iterations is limited and, thus, the decoding operation is stopped after several iterations, whether the data is error-free or not. This is one reason that it is important that the quality of the decoded data improve from one iteration to the next.
In addition to turbo-decoding, many practical systems use additional independent means to minimize the errors in the transmitted data. For example, in 3GPP standards for wireless communication, cyclic redundancy check (CRC) blocks are added to the transmitted data. The length of the CRC blocks may or may not coincide with the length of the turbo-decoder block. Each CRC block is attached at the end of a respective transport block. Subsequently, several such CRC blocks can be concatenated to form a code block.
Conventional methods of turbo-decoding only deal with cases where the coded frame contains a single CRC code block (where “block” means the original data and the CRC generated for the original data). For example, a device disclosed in U.S. Pat. No. 5,761,248, the disclosure of which is incorporated herein by reference for all it teaches, discloses a method for determining an adaptive abort criteria. Further, a device disclosed in U.S. Pat. No. 6,182,261, the disclosure of which is also incorporated herein by reference for all it teaches, stops the iteration process if the CRC check passes. Neither of the devices disclosed in these two patents addresses the situation where there is a plurality of CRC code words in the same coded data frame. Furthermore, the two patents mentioned do not describe how to handle, or even recognize, the situation where some of the CRC code words pass the CRC check and some of the CRC code words fail the CRC check.
MUX (13) alternately selects bits from the original data, i.e., the “Systematic” bits, and the two parity bits output respectively from the first and second RSC encoding blocks. Thus, for every bit at the input to the Turbo Encoder (1), each ofthree different bits are alternately output—a Systematic Bit, a First Parity Bit and a Second Parity Bit.
In particular, a Systematic Bit and a First Parity Bit enter Decoder 1 (20). Additionally, extrinsic information for each bit, as calculated during the previous iteration, enters Decoder 1 (20) as well. The output of Decoder 1 (20) comprises a new extrinsic value for each input bit. For example, a “extrinsic” number comprises a digital representation of how likely the original data bit was a “0” or how likely the original data bit was a “1”. The range of values for the extrinsic number can be, for example, from −7 to +7. A “strong zero”, or in other words, a representation that the bit is most likely a “0” can be shown as −7 and a representation that the bit is most likely a “1” can be shown as +7, with varying degrees of “extrinsic” being represented by the numbers between −7 and +7.
As shown in
The output from De-Interleaver (23) also feeds a Hard Decision block (24), which translates the extrinsic information output from De-Interleaver (23) into a hard decision (i.e., “1” or “0”). The Hard Decision output is then presented to a CRC Check block (25), which uses a standard CRC checking algorithm to determine if the output sequence of hard decisions from Hard Decision block (24) is correct. If the CRC test fails, nothing can be done, except at the system level where it may be possible to request a complete retransmission of the data.
An entirely turbo-decoded block of data must be written into the Interleaver (21) before the output of Interleaver (21) can be fed to Decoder 2 (22). The reason for this is because the interleaving process is defined so that all (or almost all) input bits must be stored before the output bits can be generated. This is true for De-Interleaver (23) as well. Accordingly, the circuit works in two distinct phases, as follows:
Phase 1: through Decoder 1 (20) and output from Interleaver (21). In this phase, the data is in interleaved order and therefore the CRC cannot be checked.
Phase 2: through Decoder 2 (22) and out from De-Interleaver (23). Only in this phase can the CRC be checked.
Also, the length of a CRC block does not necessarily coincide with the length of the Turbo Decoder block.
In view of the issues raised above in regard to inefficient and/or ineffective decoding, an object of the present invention is to provide a system and method of error correcting relative to data transmission in which time requirements and power consumption in the decode process are reduced.
To achieve the above-mentioned object of the present invention, a system in accordance with one embodiment of the present invention includes most of the functional structures required to carry out a conventional turbo-decode process, a process that is fully described in the references incorporated hereinabove. In addition to the functional structures traditionally required, however, a system in accordance with the present embodiment includes several novel functional structures for reducing both the time required to carry out the turbo-decode process as well as the power consumed by the system in the process.
For example, the present embodiment includes a turbo-coding system including a turbo-encoder for encoding data to be transmitted and a turbo-decoder for receiving the transmitted data as turbo code blocks and decoding the data. The turbo-decoder also verifies that the reliability of the received data is above a certain degree of reliability. The turbo-decoder in this embodiment includes a CRC checking unit that determines whether each code block either passes or fails a CRC checking algorithm. If the code block passes the CRC checking algorithm, the results of the CRC checking algorithm are fed back to the iterative turbo-decoding process. As a result of the information fed back from the CRC checking algorithm, the turbo decoding process is able to increase the speed by which errors are corrected and, thus, both time and energy are conserved.
In accordance with the present embodiment, when a code block passes a respective CRC check, an address calculator and a range check unit are activated. Also, until the CRC check unit determines that a particular CRC block passes the CRC check algorithm, an extrinsic saturation unit is controlled to be OFF. While in the OFF condition, the extrinsic saturation unit acts as a conduit for interleaved data leaving the interleaver unit, i.e., so-called ‘soft’ data, and the second decoder.
However, as soon as the CRC check unit determines a ‘pass’ result for a particular code block, the range check unit, which has been activated by the CRC check unit, turns ON the extrinsic saturation unit for all data corresponding to the code block that resulted in the passed CRC. The extrinsic saturation unit, thus, ‘saturates’ the extrinsic data for each of the bits of data that made up the passed CRC code block. In other words, the data into the turbo-decoder's second decoder is set at the data's corresponding maximal soft value.
The second decoder operates as usual on the saturated data and after decoding the data, passes the decoded data along to the De-Interleaver where the original order of the data is retrieved. Experience has shown that the saturated data, resulting from the code block that passes the CRC, will improve the performance of the of the Turbo decoder in such a way that bits that do not correspond to the passing CRC will get better soft symbols, and as a result, the turbo decode process is halted sooner than it would have been had the CRC check result not been injected into the intermediate processing of the overall turbo-decode process.
The invention as well as embodiments and advantages thereof are described below in greater detail, by way of example, with reference to the drawings in which:
In particular, upon determining a ‘passed’ CRC check, Extrinsic Saturation block (32) forces “saturation” (i.e., maximal soft value) of the extrinsic value on the data stream. The extrinsic value comprises words of several bits. The number of bits used for the extrinsic value is arbitrary, however, typically 3-5 bits of data are used to represent extrinsic information. In the example discussed above, where the extrinsic information value ranged from −7 to +7, four bits were used; one bit to represent the sign of the number (− or +) and three bits to represent the actual value (0-7). Thus, each extrinsic “word” can be regarded as the binary representation of an integer, either positive or negative. These integers represent extrinsic values. The largest integer that can be represented with the given number of bits is the ‘maximal soft value’.
From the time the first turbo-decode iteration begins, and until a first CRC check results in a “pass” condition, the extrinsic saturation block (32) is in the OFF position, that is, it does not saturate the extrinsic value but it does transfer the soft symbols from its input to its output. In other words, until the CRC check unit (36) determines a “pass” for a particular block, circuit (100) operates exactly like the decoder circuit depicted in FIG. 2.
When the CRC check unit (36) determines its first ‘passed’ CRC block, the address calculation unit (37) and the range check unit (38) are activated. The range check unit (38), in turn, turns the Extrinsic Saturation unit (32) ON for all symbols that correspond to the present CRC block. Extrinsic Saturation unit (32) subsequently sets the reliability value of the instant bit to its respective maximal soft value, provided that the bit belongs to a sub-frame with a “passed” CRC. In other words, after the CRC makes its first ‘passed’ determination, the Extrinsic Saturation unit (32) is controlled to set each extrinsic value for the corresponding data block (i.e., the block that passed its CRC) to its maximum value. That is, relatively weak “0”s and “1”s are converted into very strong “0s” and “1”s, respectively.
In accordance with the present embodiment, half an iteration is consumed by the time the CRC makes its determination. In other words, half an iteration is wasted because after the first phase, as the data is output from Interleaver (31), the data is still in interleaved order and, thus, CRC checking cannot be performed yet. The CRC check unit (36) determines whether a CRC block ‘passes’ the CRC algorithm only after all symbols are output from the De-Interleaver (34), i.e., after the second phase, when the data has already been fed back to and decoded by Decoder 1 (30), and stored in Interleaver (31). Therefore, it is only possible to force a hard-decision at the next half-iteration, e.g., when symbols are transferred from Interleaver (31) to Decoder 2 (33), and into De-Interleaver (34). Accordingly, a mechanism is required to keep track of the data and its corresponding interleaved ordering such that when the CRC determines a ‘pass’ for a particular block immediately following the second phase, the corresponding data that has been fed back to the first phase and, thus interleaved, can be retrieved properly and operated on by the Extrinsic Saturation unit (32). For this purpose, Address Calculator (37) and Range Check unit (38) are provided.
When symbols are read from Interleaver (31), the addressing order is such that interleaving occurs, e.g., symbols that were written in rows are read in columns. Any type of interleaving algorithm can be used, e.g., a square interleaver, a triangle interleaver or a random interleaver. Those skilled in the art would know how to build and/or use an Interleaver and the present invention is not limited to any one type of interleaving algorithm. As a result of interleaving, the range of addresses for a certain CRC block is not continuous when read from Interleaver (31). The addresses corresponding to the data in Interleaver (31) are calculated by the Address Calculator (37). For example, the structure of the interleaved block output from Interleaver (31) is such that the information is written in consecutive addresses, and interleaving is performed by controlling the order of the read operation.
A description in accordance with the present embodiment in regard to the Address Calculator (37) and the Range Check unit (38) is now provided.
A task of the Address Calculator (37) is to generate the original address of the location into which the data was originally written into Interleaver (31). It is necessary to keep track of this information because there is a difference in the order of the data being read from the Interleaver (31), as compared to the order as the data is written into the Interleaver (31). The difference in addresses is a function of the interleaver algorithm used.
A task of the Range Check unit (38) is to identify which of the data blocks the current sample corresponds to and determine whether the current sample has passed the CRC check in the previous round and should now be changed to the maximum value, as discussed above, or maintained at its current value. The operation of the Range Check unit (38) is, therefore, also specific to the algorithm used in the interleaving, and also specific to the size of the code blocks. For example, consider a simple square interleaver having blocks of 4 bits each, and each block having a corresponding CRC value comprising a single bit. If it is assumed that there are four such blocks to transfer, the four data bits in the first block are numbered 1-4 and the CRC bit is numbered 5. The data bits in the second block are numbered 6-9 with the corresponding CRC bit numbered 10, and so on for the third and fourth blocks. Therefore, the numbering of the bits for all four blocks (20 bits) is 1, 2, 3, 4, . . . 19, 20. When the data is written into the Interleaver (31), the data structure shown in
As mentioned above, a task of the Range Check unit (38) is to determine whether the current data belongs to a data block that has ‘passed’ the CRC check. For example, considering the second data sample read from Interleaver (31), according to the address generator table of
Thus, the Range Check unit (38) may comprise a table, such as the table shown in
In accordance with the present embodiment, the read address for each symbol read from Interleaver (31) corresponds to the order in which that symbol was written into the block. That is, when the n-th symbol, that was previously written into De-Interleaver (34), is read out of De-Interleaver (34), the read address is ‘n’. The read address can be used by the Range Check unit (38) to determine whether or not each symbol belongs to a CRC group which has already passed CRC check.
Thus, a real-time system results in which the turbo-decoding process is improved, by using the CRC result in the iterative turbo-decoding process, and power consumption is conserved since the iterative process is halted and a hard decision is forced for at least some of the decoded data as soon as the CRC determines a ‘pass’.
However, for applications in which the latency required to update the data and rerun the turbo decoder process can be tolerated, an improvement over the conventional turbo-decoder process shown in
One of the advantages of the present invention over conventional turbo-code methods is that additional information, e.g., CRC data, is used throughout the turbo-decoding process. Using the additional information results in a smaller number of iterations, statistically, and, hence, lowers the overall power consumption of the encoding/decoding circuits.
The above description of the preferred embodiments has been given by way of example. From the disclosure given, those skilled in the art will not only understand the present invention and its attendant advantages, but will also find apparent various changes and modifications to the structures and methods disclosed. It is sought, therefore, to cover all such changes and modifications as fall within the spirit and scope of the invention, as defined by the appended claims, and equivalents thereof.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5446747 *||Apr 16, 1992||Aug 29, 1995||France Telecom||Error-correction coding method with at least two systematic convolutional codings in parallel, corresponding iterative decoding method, decoding module and decoder|
|US5761248||Jul 19, 1996||Jun 2, 1998||Siemens Aktiengesellschaft||Method and arrangement for determining an adaptive abort criterion in iterative decoding of multi-dimensionally coded information|
|US6182261 *||Jul 9, 1999||Jan 30, 2001||Qualcomm Incorporated||Efficient iterative decoding|
|US6487694 *||Dec 20, 1999||Nov 26, 2002||Hitachi America, Ltd.||Method and apparatus for turbo-code decoding a convolution encoded data frame using symbol-by-symbol traceback and HR-SOVA|
|US6526531 *||Mar 22, 2000||Feb 25, 2003||Agere Systems Inc.||Threshold detection for early termination of iterative decoding|
|US6614858 *||Feb 17, 2000||Sep 2, 2003||Agere Systems Inc.||Limiting range of extrinsic information for iterative decoding|
|US6686853 *||Aug 15, 2002||Feb 3, 2004||Broadcom Corporation||Method and apparatus for iterative decoding|
|1||C. Berrou and A. Glavieux, Turbo-Codes: General Principles and Applications, Audio and Video Digital Radio Broadcasting Systems and Techniques, 1994, pp. 215-226, Ecole Nationale Supérieure des Télécommunications de Bretagne, BP 832-29285, Brest Cedex, France.|
|2||C. Berrou, A. Glavieux and P. Thitimajshima, Near Shannon Limit Error-Correcting Coding and Decoding: Turbo-Codes (1), 1993, pp. 1064-1070, Ecole Nationale Supérieure des Télécommunications de Bretagne, BP 832-29285, Brest Cedex, France.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7272771 *||Oct 30, 2003||Sep 18, 2007||Nokia Corporation||Noise and quality detector for use with turbo coded signals|
|US7383484 *||Mar 12, 2004||Jun 3, 2008||Seagate Technology Llc||Cyclic redundancy check based message passing in turbo product code decoding|
|US7441175 *||Mar 12, 2004||Oct 21, 2008||Seagate Technology Llc||Turbo product code implementation and decoding termination method and apparatus|
|US7840873||Dec 13, 2006||Nov 23, 2010||Globalfoundries Inc.||Partial CRC insertion in data packets for early forwarding|
|US7881303||Dec 13, 2006||Feb 1, 2011||GlobalFoundries, Inc.||Command packet packing to mitigate CRC overhead|
|US8051239||Jun 4, 2007||Nov 1, 2011||Nokia Corporation||Multiple access for parallel turbo decoder|
|US8074155 *||Dec 6, 2011||Broadcom Corporation||Tail-biting turbo coding to accommodate any information and/or interleaver block size|
|US8090896||Jul 3, 2008||Jan 3, 2012||Nokia Corporation||Address generation for multiple access of memory|
|US8327234 *||Feb 27, 2009||Dec 4, 2012||Research In Motion Limited||Code block reordering prior to forward error correction decoding based on predicted code block reliability|
|US8364916 *||Jun 17, 2009||Jan 29, 2013||Zte Corporation||Method and apparatus for implementing interleaving and de-interleaving at second time|
|US20050097431 *||Oct 30, 2003||May 5, 2005||Nokia Corporation||Noise and quality detector for use with turbo coded signals|
|US20050204255 *||Mar 12, 2004||Sep 15, 2005||Nan-Hsiung Yeh||Cyclic redundancy check based message passing in Turbo Product Code decoding|
|US20050204256 *||Mar 12, 2004||Sep 15, 2005||Nan-Hsiung Yeh||Turbo Product Code implementation and decoding termination method and apparatus|
|US20080301383 *||Jun 4, 2007||Dec 4, 2008||Nokia Corporation||Multiple access for parallel turbo decoder|
|US20100005221 *||Jan 7, 2010||Nokia Corporation||Address generation for multiple access of memory|
|US20100031125 *||Feb 4, 2010||Broadcom Corporation||Tail-biting turbo coding to accommodate any information and/or interleaver block size|
|US20100223534 *||Feb 27, 2009||Sep 2, 2010||Research In Motion Limited||Code Block Reordering Prior to Forward Error Correction Decoding Based On Predicted Code Block Reliability|
|US20110087849 *||Jun 17, 2009||Apr 14, 2011||Zte Corporation||Method and apparatus for implementing interleaving and de-interleaving at second time|
|US20130013976 *||Sep 13, 2012||Jan 10, 2013||Research In Motion Limited||Code block reordering prior to forward error correction decoding based on predicted code block reliability|
|USRE44487||Sep 22, 2011||Sep 10, 2013||Globalfoundries Inc.||Command packet packing to mitigate CRC overhead|
|U.S. Classification||714/794, 714/758|
|International Classification||H03M13/29, H03M13/09, H03M13/27|
|Cooperative Classification||H03M13/2975, H03M13/09|
|Jul 16, 2002||AS||Assignment|
Owner name: 3G.COM, INC., DELAWARE
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:LAVI, YOAV;BONER, ALON;REEL/FRAME:013096/0777;SIGNING DATES FROM 20020702 TO 20020703
|Apr 6, 2009||REMI||Maintenance fee reminder mailed|
|Sep 27, 2009||LAPS||Lapse for failure to pay maintenance fees|
|Nov 17, 2009||FP||Expired due to failure to pay maintenance fee|
Effective date: 20090927